|Home | About | Journals | Submit | Contact Us | Français|
The formation, maintenance and plasticity of neural circuits rely upon a complex interplay between progressive and regressive events. Increasingly, new functions are being identified for axon guidance molecules in the dynamic processes that occur within the embryonic and adult nervous system. The magnitude, duration, and spatial activity of axon guidance molecule signaling are precisely regulated by a variety of molecular mechanisms. Here we focus on recent progress in understanding the role of protease-mediated cleavage of guidance factors required for directional axon growth, with a particular emphasis on the role of metalloprotease and γ-secretase. Since axon guidance molecules have also been linked to neural degeneration and regeneration in adults, studies of guidance receptor proteolysis are beginning to define new relationships between neurodevelopment and neurodegeneration. These findings raise the possibility that the signaling checkpoints controlled by proteases could be useful targets to enhance regeneration.
More than 2000 years ago the Chinese advanced the concept of Yin and Yang to help explain the progressive and regressive forces that operate during life. As depicted in the Taiji diagram, Yin and Yang are interdependent, interconnected and transformable. Today, this ancient theory still provides a useful conceptual framework for viewing the dynamic relationship between progressive and regressive events during neural development and maintenance. Here we describe how this principle applies to multiple organizational levels of the nervous system: from circuits, to cells, to molecules (Figure 1A).
Neural circuit formation involves many progressive events including neural stem cell proliferation, axon and dendrite outgrowth, and synapse formation. Later in development, however, regressive events such as cell death, axon pruning, and synapse elimination further refine the precise pattern of connectivity needed for proper function of the mature circuitry. Neuron death and synapse loss also occur under pathophysiological conditions such as Amyotrophic lateral sclerosis and Alzheimer’s disease, for example (Vanderhaeghen and Cheng, 2010). Unfortunately, the counter-forces that might offset the degeneration of neural circuits seem to be far less robust in the adult than the embryonic nervous system of higher vertebrates, creating a major clinical challenge (Giger et al., 2010).
Progressive and regressive events also apply to the attractive and repulsive forces that guide growing axons (O’Donnell et al., 2009). Some cellular targets express attractive cues, which promote the assembly of cytoskeletal networks within growth cones, leading to axonal turning and extension; whereas other targets express repulsive cues that cause cytoskeleton disassembly. Interestingly, the signaling pathways that cause axon attraction and repulsion are transformable when levels of cyclic nucleotides and Ca2+ are altered (Hong et al., 2000; Hopker et al., 1999; Nishiyama et al., 2003). In fact, the responsiveness of axons to guidance signals often changes over their course of growth. Long axons typically navigate using a series of intermediate targets. For each intermediate target, the axon is first attracted then switches its response upon arrival and becomes repelled, allowing it to move on to the next leg of its journey (Tessier-Lavigne and Goodman, 1996; Yu and Bargmann, 2001).
At a molecular level, the synthesis and degradation of proteins can likewise be viewed as progressive and regressive processes. Accordingly, it is easy to understand how the synthesis of new proteins is critical for cell proliferation, the specification of neuronal identity, and axonal extension. Counter-intuitively, recent studies have found that regressive factors like proteases are also critical for the assembly of neural circuits. Recently, an excellent review from Binglol and Sheng has covered the exciting progress on proteolytic regulation of synaptic proteins in neural plasticity (Bingol and Sheng, 2011). In this review we focus on the role of membrane-associated proteases that influence axon growth and guidance. We describe new findings on the sequential cleavage of axon guidance receptors by metalloproteases and γ-secretases and speculate on how this provides an additional layer of regulation to diversify the functions of guidance receptors as well as enhance the fidelity of axon navigation. We conclude by describing how protease-dependent modulation of neural growth may represent a form of plasticity that can be harnessed for neural regeneration and repair (Figure 1B).
The wiring of neural networks relies on the coordination of two separate events: the precise presentation of guidance signals and the correct receipt and processing of these signals. Considerable progress has been made in identifying extracellular cues that influence axonal growth cone dynamics, including members of the four classic guidance cue families: Netrins, Slits, Semaphorins (Semas) and Ephrins; and their respective neuronal receptors: DCC/UNC5, Robo, Neuropilin/Plexin and Ephs, as signal-receipt elements (Dickson, 2002). A number of mechanisms have been identified to ensure the correct presentation and receipt of guidance signals, including regulated endocytosis, control of receptor trafficking, receptor compartmentalization within the plasma membrane, localized mRNA transport, and regulated translation (Brittis et al., 2002; O’Donnell et al., 2009; Tcherkezian et al., 2010). Notably, emerging evidence from invertebrate and vertebrate studies highlight the important role of regulated proteolysis in modulating the spatial and temporal pattern of guidance receptors and cues during the assembly of neural circuits (O’Donnell et al., 2009).
The human genome encodes over 500 proteases, representing ~1.5% of the protein-coding genes (Puente et al., 2003). They are divided into six families based on the nucleophile used to break peptide bonds: serine, threonine, cysteine, aspartic acid, metallo, and glutamic acid. These enzymes display exquisite substrate specificity and are associated with a wide range of biological processes from catabolism of protein for nutrition, to protein quality control, to protein maturation (Fujinaga et al., 2004; Hooper, 2002). Notably, proteases are also important for modulating the kinetics and quality of signal produced by receptor-ligand interactions (Hooper, 2002). They can control the (1) spatial distribution and levels of proteins, (2) activation of receptors, (3) duration of signaling, and (4) downstream pathway selection. In the case of Notch, receptor cleavage is necessary to initiate intracellular signaling (Selkoe and Kopan, 2003). In contrast, signaling from Plexin-A1 receptors is terminated by protease cleavage, possibly to ensure that rapid biological processes such as axon growth are coupled to potent but short lived signal transduction pathways (Nawabi et al., 2010).
An initial role for proteolysis in axon guidance came from studies that showed growth cones secrete some proteases such as metalloproteases, hypothesized to chew up the extracellular matrix, and thereby clear a passage for axons (Krystosek and Seeds, 1981; Muir, 1994; Schlosshauer et al., 1990). Metalloproteases represent a large family of Zinc-dependent proteolytic enzymes including secreted (ADAMTSs, MMPs and Pappalysins), membrane-bound (ACEs, ADAMs and MT-MMPs), and cytosolic proteases (Insulysin, Neprilysins and THOP1) (Apte, 2009; Boldt et al., 2001; Hadler-Olsen et al., 2011; Imai et al., 2007; Malito et al., 2008; Shrimpton et al., 2002; Yong et al., 2001). Recent studies suggest that MMPs (matrix metalloproteases) and ADAMs (a disintegrin and metalloproteinases) control axonal growth directly by cleaving axon guidance receptors and ligands.
Pioneering studies by Galko and Tessier-Lavigne revealed that metalloprotease was involved in the ectodomain shedding of DCC (Galko and Tessier-Lavigne, 2000). They found that blocking metalloprotease activity enhanced full-length DCC receptor levels and potentiated Netrin-induced axon outgrowth from spinal cord explants. Although the identity of the metalloprotease involved in DCC cleavage remains unknown, these data provided the first direct evidence that metalloprotease-mediated receptor cleavage modulates axonal responsiveness by regulating the number of functional axon guidance receptors on the plasma membrane (Figure 2A). Although DCC cleavage attenuates chemoattraction, there are some instances where regulated receptor proteolysis is required to activate axon guidance signaling.
Through a genetic screen for axon guidance defects in drosophila, the kuzbanian (kuz) mutant was identified with defective midline repulsion leading to inappropriate midline crossing of ipsilateral interneurons. Kuz is a single-pass ADAM family transmembrane metalloprotease (ADAM10) that is widely expressed throughout development in the drosophila central nervous system (Fambrough et al., 1996). Genetic-interaction experiments suggest that Kuz positively regulates Slit/Robo mediated repulsion, prompting questions about the molecular mechanism (Schimmelpfeng et al., 2001). This remained a puzzle until Coleman et al. discovered that Kuz cleaves the Robo receptor leading to its activation (Coleman et al., 2010). Proteolysis of Robo appears to be critical for its signaling since Robo mutations that prevent cleavage disrupt Slit-mediated repulsion of drosophila ipsilateral neurons (Figure 2B). Robo cleavage leads to the recruitment of the Sos (Son of Sevenless), which is a Ras/Rac guanine exchange factor (GEF) involved in signaling to the cytoskeleton. Further studies are needed to determine if this cleavage mechanism used to activate Robo-signaling is conserved in vertebrate neurons. Nevertheless, these findings reveal a general strategy whereby regulated receptor proteolysis can convert short-lived or weak interactions into durable signaling.
Guidance molecules like Netrin and Slit are secreted ligands, while A- and B-class Ephrins are membrane-bound proteins. Binding of Ephrins to Eph receptor-expressing neurons triggers growth cone collapse (Egea and Klein, 2007). Because Ephs and Ephrins are attached to cell membranes, this raised the question how neurons could overcome the adhesive properties of Eph-Ephrin binding in order to retract. Regulated proteolysis was found to sever the Ephrin protein, breaking the cell-cell adhesion (Figure 2C). Prior to Eph-Ephrin contact, ADAM10 constitutively associates with EphA3 receptors. Upon EphA3 interaction with Ephrin-A2, the formation of a functional EphA3/Ephrin-A5 complex creates a new molecular recognition motif for effective Ephrin-A2 cleavage by ADAM10. This breaks the molecular tether between the opposing cell surfaces and allows the internalization of EphA3/Ephrin-A5 complexes into Eph-expressing cells (Hattori et al., 2000; Janes et al., 2005).
While it is easy to imagine how metalloprotease-mediated ectodomain shedding can break adhesive interactions between cells, recent studies in drosophila have found that metalloproteases can enhance the adhesive interactions that promote axon fasciculation (Miller et al., 2008). The drosophila genome contains two matrix metalloproteases, MMP1 and MMP2. In wild-type embryos, axons of the intersegmental nerve branch b (ISNb) defasciculate from the primary ISN pathway and innervate the ventrolateral muscle (VLM) field. Misexpression of either metalloprotease disrupts the proper defasciculation of ISNb axons when they need to split apart at defined choice points. Conversely, ISNb axons in MMP mutant embryos are loosely bundled and project aberrantly within the VLM field. Similar phenotypes were also found in the guidance of the segmental nerve branch a (SNa). How could a metalloprotease potentiate the inter-axonal adhesion of motor neurons? One intriguing possibility is that MMPs regulate guidance molecules that influence axon fasciculation and defaciculation. One clue comes from the finding that axons in drosophila semaphorin-1a mutants fail to separate when they reach their targets, suggesting that Sema-1A promotes inter-axonal repulsion and defasciculation (Yu et al., 1998). Miller et al. found that decreasing the semaphorin gene dose by half (sema-1a heterozygotes) suppressed the axon fasciculation phenotype in MMP2 mutants. Despite the lack of direct evidence showing cleavage of Semaphorins or their receptors by metalloproteases, this genetic interaction suggests that MMP1 and MMP2 may modulate Semaphorin signaling and thereby control the proper timing of axon fasiculation and defasiculation required for motor axon navigation (Miller et al., 2008).
Is proteolytic cleavage of guidance molecules actively regulated during axon navigation? Although relevant evidence is lacking for metalloproteases, recent studies on Calpain-mediated receptor processing reveal regulated proteolysis as an important way for controlling guidance decisions (Figure 2D) (Nawabi et al., 2010). Calpains are calcium-dependent cysteine proteases expressed ubiquitously in both vertebrates and invertebrates. During commissural axon navigation, growth cones are held in a state desensitized to floor plate repellents in order to reach the midline; then become responsive to repellents upon contacting the floor plate. Nawabi et al. found that at the precrossing stage, commissural neurons synthesize the Neuropilin-2 and Plexin-A1 receptor subunits, but Plexin-A1 protein levels are kept low by Calpain proteolysis. During floor plate crossing, Calpain1 activity is suppressed by local NrCAM, enabling Plexin-A1 to accumulate in the growth cone, thereby sensitizing axons to Sema-3B repulsion (Figure 2D) (Nawabi et al., 2010). In the future it will be interesting to determine how NrCAM regulates Calpain activity.
Metalloproteases such as ADAM10 and MMPs cleave off the extra-cellular domain of transmembrane axon guidance ligands and receptors. What is the fate of the membrane-tethered intracellular “stubs” after metalloprotease cleavage? In the case of other known membrane-proteins, the intracellular stubs created by metalloproteases typically become substrates for γ-secretase proteases. This intra-membranous enzyme complex consists of four core members: Presenilin, Nicastrin, APH1 (Anterior Pharynx Defective 1) and PEN2 (Presenilin Enhancer 2) (Wolfe, 2006). The catalytic activity of γ-secretase is provided by Presenilin, which encodes a nine-pass transmembrane aspartyl protease (Laudon et al., 2005; Spasic et al., 2006). In mammals, there are two highly homologous presenilin genes, presenilin-1 (PS1) and presenilin-2 (PS2) (Donoviel et al., 1999). Metalloprotease cleavage often exposes masked cleavage sites that become the substrate of γ-secretase as part of a sequential proteolytic cleavage cascade within the cell membrane (Parks and Curtis, 2007). γ-secretase is known to have dozens of substrates, but two of the better known are Notch and amyloid precursor protein (APP) (Figure 3A, 3B and Table 1) (De Strooper et al., 1998; Mumm and Kopan, 2000; Parks and Curtis, 2007). Delta or Jagged binding to the Notch receptor triggers the ADAM protease to cleave Notch, releasing the extracellular domain and generating a membrane-tethered Notch stub that becomes a substrate for γ-secretase. Following γ-secretase cleavage, the Notch intracellular domain (NICD) is freed from the membrane allowing it to translocate to the nucleus where it acts as a transcriptional regulator of neurogenic genes (Figure 3B) (Selkoe and Kopan, 2003). Genetic linkage studies in familial Alzheimers disease (FAD) patients have identified more than 180 different mutations in the PS1 gene and an additional 13 in PS2 (http://www.molgen.ua.ac.be/ADMutations). FAD-linked PS mutations cause neomorphic protease activity, leading to increased production of Aβ42, the more hydrophobic and aggregation-prone peptides, compared to Aβ40 (Figure 3A) (De Strooper and Annaert, 2010). The prevailing amyloid hypothesis posits that accumulation of Aβ42 peptides triggers a pathogenic cascade, leading to neurodegeneration (Hardy and Higgins, 1992). Therefore, it is thought that inhibition of γ-secretase activity may help to lower Aβ production serving as a therapy for AD. This therapeutic approach is tempered by the finding that conditional inactivation of PS1/2 in the adult mouse brain causes progressive memory loss and neurodegeneration (Saura et al., 2004; Zhang et al., 2009), raising the possibility that γ-secretase activity is required for maintaining normal brain function (Shen and Kelleher, 2007).
Notch and APP represent only two examples of an expanding list of γ-secretase substrates that are known to undergo sequential proteolytic cleavage. Although this list includes many axon guidance molecules, the functional consequences of γ-secretase cleavage are best defined for the Netrin receptor DCC (Table 1). The extracellular domain of DCC is first cleaved by a metalloprotease to create a membrane-tethered DCC stub. Under normal conditions the DCC stub is present at low concentrations because it is rapidly cleaved by γ-secretase, releasing the intracellular domain (ICD) from the membrane (Figure 3C). In vitro studies have shown that inhibition of γ-secretase activity results in accumulation of DCC stubs in cell membranes and is correlated with enhanced neurite outgrowth in cultured neuroblastoma cells (Figure 2E) (Parent et al., 2005; Taniguchi et al., 2003).
Although PS1 and PS2 mutant mice have been studied for more than a decade, the in vivo function of γ-secretase cleavage of guidance molecules such as DCC was uncertain, perhaps because of the diversity of PS functions (Donoviel et al., 1999; Shen et al., 1997). A role for PS1 in axon guidance was first revealed in a mouse ENU mutagenesis screen to identify genes involved in embryonic motor neuron axon pathfinding (Bai et al., 2011). Bai and colleagues discovered that motor axons in the Columbus mutant grew into the spinal cord floor plate at the midline rather than exiting laterally through their normal ventral root sites and found this phenotype is caused by a mutation in the PS1 gene. Through a series of in vivo and in vitro experiments they linked this axon guidance phenotype to a defect in γ-secretase processing of DCC, causing motor neurons to inappropriately become attracted to the Netrin-1 produced by the floor plate (Bai et al., 2011).
Normally γ-secretase-mediated cleavage has two effects on receptor processing: (1) clearing receptor stubs from the membrane, and (2) releasing intracellular domain fragments (Figure 3C and 3E). The signaling ability of cleaved receptor intracellular domains has been shown in several examples. In the case of Ephrin-B reverse signaling, cleavage of Ephrin-B releases an ICD that binds Src, which disrupts its association with the inhibitory kinase Csk, allowing autophosphorylation of Src and activation of signaling (Figure 3G) (Georgakopoulos et al., 2006). Similarly, EICD, the intracellular domain fragment of EphA4 has been shown to be required for activiation of Rac signaling (Inoue et al., 2009). By analogy to Notch (Figure 3B), the DCC-ICD is considered a possible nuclear signaling intermediate because fusion to the Gal4 DNA binding domain revealed it could activate transcription using reporter assays (Taniguchi et al., 2003). Nevertheless, electroporation of DCC-ICD expression constructs into chick spinal motor neurons failed to alter motor axon projections suggesting this fragment of DCC is not involved in motor axon growth. Moreover, DCC-ICD expression failed to prevent motor axon attraction to Netrin-1 in the presence of γ-secretase inhibitors, again supporting the notion that the motor neuron phenotypes in PS1 mutants do not arise from a lack of DCC-ICDs (Bai et al., 2011). In the future it will be intriguing to explore the physiological function of intracellular guidance receptor fragments in neural development.
A number of findings support the idea that receptor stubs, particularly DCC stubs, are also potent signaling components. For example, with the accumulation of DCC stubs by γ-secretase inhibition, cAMP-dependent signaling is also increased in both neuroblastoma cells and cortical neurons (Parent et al., 2005). Overexpression of myr-UNC40 (a myristoylated form of the DCC intracellular domain that mimics the DCC stub in C. elegans causes axon growth defects by activating a series of downstream kinases (Gitai et al., 2003). Forced expression of membrane-tethered DCC stubs resistant to γ-secretase cleavage caused motor neurons to become responsive to Netrin-1 (Bai et al., 2011). Intriguingly, they found that DCC stubs seem to possess properties that are distinct from the full-length (FL) DCC receptor. Based on their model, newly generated motor neurons co-express Slit-ligands and Robo-receptors (Brose et al., 1999), leading to autocrine activation of Robo, which blocks DCC’s responsiveness to Netrin-1 thereby preventing abnormal attraction to the midline (Bai et al., 2011; Stein and Tessier-Lavigne, 2001). In this process, Robo preferentially interacts with full-length DCC receptor complexes, whereas the heterogeneous DCC stub/DCC-FL complex is freed from Robo-silencing (Figure 3E) (Bai et al., 2011). Since this new complex retains the ability to signal axonal growth and is uncoupled from Robo-silencing, motor neurons become attracted to the Netrin-expressing floor plate due to the accumulation of DCC stubs in PS1 mutants (Figure 3E). The structural basis for Robo’s silencing of full-length but not DCC stubs is unknown but would be informative for understanding this molecular switch.
Thus, the axon guidance receptor DCC is the substrate of sequential proteolysis by metalloproteases and γ-secretases, which generate cleavage products with unique properties. Notably, the function of sequential proteolytic processing could be interpreted in a highly cell-type dependent manner (Bai et al., 2011; Galko and Tessier-Lavigne, 2000). For example, pre-crossing commissural neurons are attracted to Netrin, whereas newly generated motor neurons are non-responsive to Netrin because they actively silence DCC signaling by co-expressing both Slit and Robo. The inhibition of metalloproteases enhances full-length DCC receptor levels on cell surfaces, but motor neurons seem to have adequate levels of Slit and Robo to silence the additional DCC. In commissural neurons the elevated levels of DCC produced by blocking metalloprotease activity lead to enhanced Netrin-responsiveness (Figure 2A and 3C–E).
In the future it could be interesting to explore how regulated proteolysis cooperates with other modulatory mechanisms controlling axon guidance such as endocytosis, receptor trafficking, localized mRNA transport and translation. For example, when Netrin binds to DCC, signaling is activated that triggers DCC mRNA translation within the growth cone (Tcherkezian et al., 2010), raising the possibility that DCC-receptor proteolysis also modulates signaling to the translation machinery.
Studies of DCC cleavage have begun to reveal how the kinetics, substrate specificity, and spatio-temporal distribution of proteases help to form sophisticated regulatory switches that gate how axon guidance information is interpreted by neurons. In fact, highly dynamic and extremely precise control of enzyme activity represents a common feature of all protease-pathways. Regulation of proteolytic cleavage often happens at multiple levels: expression/synthesis of the components, assembly of multi-component cleavage complexes, activation of catalytic activity, interactions with enzyme modulators, and control of the spatiotemporal distribution of the enzymes and their substrates (Antalis et al., 2010; De Strooper and Annaert, 2010; Hadler-Olsen et al., 2011; Hunt and Turner, 2009; Klein and Bischoff, 2011; Kuranaga, 2011; Otlewski et al., 2005). Here we will focus on the recent progress in understanding the regulation of γ-secreatase activity at the level of (1) its subcellular localization, (2) its enzymatic activation and deactivation, and (3) modulation of its substrate specificity. Each of these regulatory layers is described in greater detail.
Although further studies are warranted, several observations indicate that γ-secretase is dynamically localized within cell membranes and endosomes (De Strooper and Annaert, 2010). Agonists of the β2-adrenergic and δ-opioid receptors induce endocytosis and concomitantly target γ-secretase to late endosomal/lysosomal compartments, leading to increased Aβ production (Ni et al., 2006). Affinity purification of active γ-secretase complexes using a Tap-tag approach found this protease was associated with tetraspanin proteins and present in detergent-resistant raft-like microdomains (Wakabayashi et al., 2009). Interestingly, altering the levels of the tetraspanin proteins CD9 or CD81 altered γ-secretase processing of APP (Wakabayashi et al., 2009). There appears to be a tetraspanin membrane-code that regulates γ-secretase activity, based on the finding that different tetraspanins (TSPAN5 and TSPAN33) from those involved in APP processing are needed for Notch cleavage (Dunn et al., 2010). A tool that would be extremely helpful for further studies of the spatio-temporal regulation of γ-secretase is a sensitive reporter system for detecting cleaved substrates at a subcellular level.
The catalytic PS1 subunit of the γ-secretase complex is phosphorylated by several kinases, including glycogen synthase kinase 3β (GSK3β), cyclin-dependent kinase 5 (Cdk5), protein kinase A (PKA), and dual-specificity tyrosine (Y)-phosphorylation-regulated kinase 1A (Dyrk1A) (Fluhrer et al., 2004; Kirschenbaum et al., 2001a, b; Lau et al., 2002; Ryu et al., 2010). These findings raise the possibility that γ-secretase activity is regulated by extracellular signals that control these kinases. Recent findings have shown that the pro-oxidant H2O2 and inflammatory cytokine pathways (interferon-γ, interleukin-1β and tumor necrosis factor-α) can stimulate γ-secretase activity and Aβ production via JNK-dependent MAPK pathways (Liao et al., 2004; Shen et al., 2008). Similarly, Kim et al. found that phosphorylation of the Nicastrin subunit by EGF-activation of ERK1/2 reduces γ-secretase activity (Kim et al., 2006). Perhaps similar extracellular cues influence γ-secretase activity in developing neurons in order to fine tune when, where and how much axon guidance signaling occurs.
Incorporation of different proteins into the γ-secretase complex may help to control the enzymatic specificity of PS1. For example, TMP21, GPCR3 and different Aph1 isoforms have been found to modulate APP processing without changing Notch cleavage (Chen et al., 2006; Serneels et al., 2009; Thathiah et al., 2009). Likewise, He et al. recently identified the GASP protein in a ternary complex with γ-secretase and found it increased Aβ production selectively (He et al., 2010). These results support the concept that cofactors help to define the substrate specificity of the γ-secretase core enzyme complex.
Numerous regressive processes occur throughout life that refine and alter the function of neural circuits including cell death, axon pruning, and synapse reorganization (Figure 1A) (Vanderhaeghen and Cheng, 2010). Although our understanding about the role of guidance molecule proteolysis in these processes is still very fragmentary, recent studies have begun to identify novel functions for axon guidance molecules associated with neurodegeneration. For example, DCC, UNC5 and EphA4 also function as “dependence receptors” that regulate cell survival (Table 2) (Mehlen and Bredesen, 2011). In the absence of their ligands, signaling is triggered by caspase cleavage of their intracellular domains, which releases a pro-apoptotic receptor fragment or permits the exposure of death domains. Consequently, overexpression of DCC or UNC5 in cultured neuronal cells induces massive apoptosis in the absence of Netrin ligand, and depletion of Netrin triggers cell death in several classes of DCC- and/or UNC5-expressing neuronal classes (Furne et al., 2008; Llambi et al., 2001; Shi et al., 2010; Takemoto et al., 2011). Likewise, removal of Ephrin-B3 ligand triggers cell apoptosis in the adult subventricular zone where its cognate EphA4 receptor is expressed (Furne et al., 2009).
A number of axon guidance molecules are also implicated in stereotyped pruning processes in the central nervous system (Vanderhaeghen and Cheng, 2010). Mutant mouse analysis reveals that blocking Sema-3A/Plexin-A3 signaling causes hippocampo-septal pruning defects; disrupting Sema-3F, Nrp-2 or Plexin-A3/4 expression affect the pruning of the infrapyramidal bundle (IPB) and visual corticalspinal tract (CST) (Bagri et al., 2003; Faulkner et al., 2007; Low et al., 2008; Sahay et al., 2003). Similarly, Xu and Henkemeyer found that EphB/Ephrin-B reverse signaling is critical for pruning of exuberant IPB fibers (Xu and Henkemeyer, 2009). Although the role of guidance molecule proteolysis in axon pruning still remains unknown, it has been reported that BACE/γ-secretase-mediated cleavage is critical for regulating axon pruning in commissural neurons and sensory neurons (Nikolaev et al., 2009).
Recent studies highlight the important roles of guidance molecule proteolysis in regulating neuronal plasticity. Neuropsin is a serine protease uniquely positioned to facilitate stress-induced plasticity due to its high expression in the amygdala and hippocampus (Chen et al., 1995). Stress results in neuropsin-dependent cleavage of EphB2 in the amygdala causing dissociation of EphB2 from NMDA receptor, thus increasing excitatory synaptic currents and enhancing behavioral signatures of anxiety (Attwood et al., 2011). Inoue et al. also found that γ-secretase-mediated EphA4 processing regulates the morphogenesis of dendritic spines. This EphA4-cleavage is disrupted by FAD mutations in PS1, raising the possibility that abnormal processing of EphA4 might contribute to AD pathogenesis or affect the maintenance and repair of neuronal circuits (Inoue et al., 2009). Along this line, future studies on protease-mediated regulation of guidance signaling pathways could provide new insight into the molecular relationships between neural development and degeneration (Figure 2B).
Neurodegenerative disease or injury often results in permanent neurological deficits due to the inability of mature CNS axons to regenerate, creating a major clinical challenge. The growth inhibitory nature of adult CNS tissue and reduced axon growth ability of adult neurons are two major barriers to regenerating axonal connections (Giger et al., 2010; Silver and Miller, 2004; Yiu and He, 2006). Several proteases have been linked to axon regeneration including metalloprotease, Calpain, BACE1, chondroitinase ABC, among which matrix metalloprotease is the best-characterized (Alilain et al., 2011; Farah et al., 2011; Spira et al., 2001; Yong, 2005).
Many studies have demonstrated that MMPs facilitate axonal regeneration in the mammalian PNS (Heine et al., 2004; Kobayashi et al., 2008; Shubayev and Myers, 2004; Zuo et al., 1998), and the CNS of lower vertebrates (Chernoff et al., 2000). Researchers also found that increased MMP expression after mammalian CNS injury is correlated with areas of increased axonal outgrowth and the subsequent enhancement of functional recovery (Ahmed et al., 2005; Duchossoy et al., 2001; Hsu et al., 2006). Mice deficient in MMP-2 display fewer serotonergic fibers caudal to the injury site and significantly reduced motor recovery compared to wild-type mice after a contusive spinal cord injury (SCI) (Hsu et al., 2006). Mechanistically, MMPs could contribute to axon regeneration in multiple ways, including (1) degrading inhibitory extracellular molecules (Belien et al., 1999; Imai et al., 1994; Muir et al., 2002; Siri et al., 1995; Turk et al., 2001), (2) clearing inhibitory cellular and matrix debris (Franzen et al., 1998; Lazarov-Spiegler et al., 1996; Rapalino et al., 1998; Rosenberg et al., 1998; Yong et al., 2001), and (3) providing trophic support to regenerating axons by degrading the ECM and releasing sequestered growth factors like bFGF (Mott and Werb, 2004). On the other hand, these positive effects are tempered by MMPs detrimental effects on mediating early secondary pathogenesis after SCI like inflammation and glial scar formation (Popovich and Longbrake, 2008; Silver and Miller, 2004). For example, mice that were treated with MMP inhibitor from 3 hours to 3 days after injury had less disruption of the blood-spinal cord barrier, fewer infiltrating inflammatory neutrophils within the spinal cord and significant locomotor recovery compared to the vehicle controls (Noble et al., 2002). Inhibition of MMP-9 release from macrophages with the use of multipotent adult progenitor cells (MAPCs) or blocking MMP-9 activity by inhibitors effectively prevent macrophage-mediated axonal retraction from the injury site (Busch et al., 2011; Busch et al., 2009). MMP-9 can also facilitate astrocyte migration and contribute to the formation of a glial scar in the injured spinal cord (Hsu et al., 2008). In addition, acute inhibition of MMP-9 after SCI induced proliferation of NG2+ cells, allowed for successful oligodendrocyte maturation and remyelination, and improved functional recovery (Liu and Shubayev, 2011). Consequently, these findings highlight the important role of MMPs in dynamically modulating the local environment of injury sites after SCI, suggesting that spatial and temporal control of individual protease activity could be useful for therapeutic applications.
Numerous axon guidance molecules including Semaphorins, Ephrins, Wnts, Slits, and Netrins become upregulated in the adult CNS following injury (Giger et al., 2010). These factors have received much attention in regenerative studies because they are candidates for modulating the growth of axons in adults. Blocking EphA4-signaling with an infused peptide antagonist enhances sprouting of corticospinal axons rostral to the injury site, but is insufficient to promote axonal regeneration across the lesion (Fabes et al., 2007). Likewise, inhibiting receptor binding of Sema-3A by a small compound (SM-216289) accelerates axon olfactory nerve regeneration and promotes serotonergic axon growth after spinal cord injury, but fails to enhance corticospinal or ascending sensory axon growth (Kaneko et al., 2006; Kikuchi et al., 2003). This is consistent with the finding that mice deficient in the receptors for class 3 semaphorins, Plexin-A3 and Plexin-A4, fail to regenerate serotinergic or corticospinal axons after a transection (Lee et al., 2010). Thus, efforts to modulate known axon guidance signaling pathways to promote axonal regeneration have met with limited success to date, but remain a promising avenue to explore for complementing other methods to promote regeneration.
In principle, targeting receptor proteolysis might provide a novel route for modulating intrinsic axonal responsiveness of adult CNS neurons (Figure 1B). The lessons learned from developmental studies of guidance receptor proteolysis suggest this strategy might be useful for (1) broadly reducing the sensitivity of inhibitory receptors, (2) increasing the sensitivity of growth promoting/attractive receptors, or even (3) switching axonal responsiveness to environmental guidance molecules from repulsion to attraction. These effects could be achieved using specific protease inhibitors or overexpression of receptor fragments like DCC stubs with potent attractive signaling activity.
The development, maintenance, and repair of the nervous system are delicately balanced between progressive and regressive events. Neural wiring, axon attraction, and local protein translation can be offset by neurodegeneration, axon repulsion, and proteolysis. These Ying and Yang events are interdependent, interconnected and transformable (Figure 1A). Increasingly, axon guidance receptor signaling has gained attention in the context of development, degeneration, and regeneration (Figure 1B and Table 2). Here, we reviewed recent progresses on our understanding of axon guidance factor proteolysis and the role that cleavage plays in transforming the activity of these important signaling proteins. While this appears to be a promising direction, many important questions remain such as whether developmental wiring defects predispose individuals to particular neurodegenerative diseases? Conversely, it remains unclear whether neurodegenerative disease-related genes/mutations influence early neural development. Can regeneration be enhanced by modulating the intrinsic receptor signaling of injured axons? We believe that studies on protease-mediated axon guidance molecule processing will provide important clues for these questions, and that the manipulation of individual proteases with high substrate-specificity might serve as clinically relevant targets to enhance regeneration.
We would like to thank Dr. Jerry Sliver and Dr. Veronica Shubayev for critical reading of the manuscript; and Jamie Simon for assistance with illustrations. We are also grateful to Dario Bonanomi, Onanong Chivatakarn and other members of the Pfaff lab for advice and discussions. G.B. is supported by the Howard Hughes Medical Institute and Pioneer foundation, and S.L.P. is a Howard Hughes Medical Institute Investigator. Research on axon guidance in the lab is supported by NINDS grants NS054172 and NS037116.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.